DARPA ULTRALOG Final Report - Industrial and Manufacturing ...
DARPA ULTRALOG Final Report - Industrial and Manufacturing ...
DARPA ULTRALOG Final Report - Industrial and Manufacturing ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Manuscript for IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS 8<br />
algorithm selection (v) <strong>and</strong> resource allocation (w) as in (2). As stated earlier, we design a<br />
scalable control mechanism to achieve the objective in the framework of MPC by building a<br />
mathematical programming model <strong>and</strong> decentralizing it.<br />
arg max<br />
v,w<br />
e<br />
i<br />
∑∑<br />
i∈ A d = 1<br />
v<br />
d<br />
i<br />
− CCT(T )<br />
(2)<br />
3. Mathematical programming model<br />
The mathematical programming model is essentially a scheduling problem formulation. There<br />
are a variety of formulations <strong>and</strong> algorithms available for diverse scheduling problems in the<br />
context of multiprocessor, manufacturing, <strong>and</strong> project management. In general, a scheduling<br />
problem is allocating limited resources to a set of tasks to optimize a specific objective. One<br />
widely studied objective is completion time (also called makespan in the manufacturing<br />
literature) as the problem we have considered. Though it is not easy to find a problem exactly<br />
same as ours, it is possible to convert our problem into one of the scheduling problems. For<br />
example, in job shop, there are a set of jobs <strong>and</strong> a set of machines. Each job has a set of serial<br />
operations <strong>and</strong> each operation should be processed on a specific machine. A job shop scheduling<br />
problem is sequencing the operations in each machine by satisfying a set of job precedence<br />
constraints such that the completion time is minimized. When we assign a value mode to each<br />
task, our problem can be exactly transformed into a job shop scheduling problem. However,<br />
scheduling problems are in general intractable. Though the job shop scheduling problem is<br />
polynomially solvable when there are two machines <strong>and</strong> each job has two operations, it becomes<br />
NP-hard on the number of jobs even if the number of machines or operations is more than two<br />
[19][20]. Considering that the task flow structure of our networks is arbitrary, our scheduling